A Closer Look At Deep Policy Gradients

Abstract

We study how the behavior of deep policy gradient algorithms reflects the conceptual framework motivating their development. To this end, we propose a fine-grained analysis of state-of-the-art methods based on key elements of this framework: gradient estimation, value prediction, and optimization landscapes. Our results show that the behavior of deep policy gradient algorithms often deviates from what their motivating framework would predict: the surrogate objective does not match the true reward landscape, learned value estimators fail to fit the true value function, and gradient estimates poorly correlate with the "true" gradient. The mismatch between predicted and empirical behavior we uncover highlights our poor understanding of current methods, and indicates the need to move beyond current benchmark-centric evaluation methods.

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyilyas2018a

Related papers

• Where Did My Optimum Go?: An Empirical Analysis Of Gradient Descent Optimization In Policy Gradient Methods (2018)0.00
• The Definitive Guide To Policy Gradients In Deep Reinforcement Learning: Theory, Algorithms And Implementations (2024)3.29
• An Empirical Analysis Of Measure-valued Derivatives For Policy Gradients (2021)0.00
• Learning Value Functions In Deep Policy Gradients Using Residual Variance (2020)0.00
• Implementation Matters In Deep Policy Gradients: A Case Study On PPO And TRPO (2020)0.00
• An Analysis Of Measure-valued Derivatives For Policy Gradients (2022)2.26
• Behind The Myth Of Exploration In Policy Gradients (2024)0.00
• Identifying Policy Gradient Subspaces (2024)0.00